Aims. To assess in current injecting heroin and amphetamine users to what extent their history of injecting represents a transition from amphetamine to heroin. Design. Cross-sectional survey. Setting. Sydney, Australia. Participants. One hundred and fifty-one primary heroin injectors and 145 primary amphetamine injectors recruited through advertisements, needle exchanges, methadone maintenance clinics and by word of mouth. Findings. Six major transition patterns were detected: heroin-heroin (n = 61), amphetamines-heroin (n = 60), heroin-amphetamines-heroin (n = 30), amphetamines-amphetamines (n = 80), amphetamines-heroin-amphetamines (n = 41) and heroin-amphetamines (n = 24). A logistic regression analysis predicting presence or absence of a transition from the original primary drug indicated that length of injecting career, years of education and original drug injected were independent predictors. Thus, the longer the injecting career, the greater the likelihood of a transition. If the original drug injected was amphetamine, the greater the likelihood of transition; and the more prior years of education, the lower the chances of a transition. Conclusions. While there was a small preponderance of movement from primary amphetamine injecting to primary heroin injecting, there was also movement in the other direction. Heroin use is not necessarily a stable endpoint for injecting careers.